Behavior of Speech and DTMF inputs (STT call module)

When using the Speech to Text module, customers can mix speech and dtmf inputs (if this feature is enabled). For your service, it might be relevant to log how a customer made such an input. In order to achieve this, you can use a set of variables to find out what the input was

• {app.speechToText.myVar} = returns the final result
• {app.speechToText.myVar.transcript} = includes e.g. digits which did not match the minimum or maximum number of inputs
• {app.speechToText.myVar.input_type} = returns whether the input was voice or dtmf
• {app.speechToText.myVar.confidence} = returns the confidence we had in the customer's input
• {app.speechToText.myVar.valid} = checks whether the input was valid

There are some general rules when it comes to mixing speech and dtmf:

• Whenever a customer starts typing an input via the dialpad, the audio channel is closed
  • This means, even if the customer starts saying something, this will not be recorded by babelforce
  • Also, if a customer started saying something and then began to type, only the typed input will be kept
  • This will be reflected in the type which will then be dtmf
• All dtmf inputs always have a confidence of 1

Numeric scenarios

Below, we give an example test row of expected behavior as the system is setup as of first quarter of 2023:

• Speech model = Command and search
• DTMF = true
• Numeric = true
• Min 2
• Max 5
• barge-In = true